MLOps
Designing evals you'll actually run
Why offline evals beat vibes, and how to build a suite your team trusts.
Blog
Notes from the workshop
Practical writing on building, shipping and operating web and AI software, in plain language.
Latest articles
LLM
RAG in production: the unglamorous parts
Chunking, freshness, citations and the failure modes nobody demos.
Cost
Cutting inference cost by 40% without losing quality
Routing, caching and distillation strategies that paid off.
Agents
Agents that don't go rogue
Guardrails, tool scoping and human-in-the-loop patterns that work.
Data
A pragmatic data platform for small teams
How to get 80% of a lakehouse with 20% of the complexity.
Web
Shipping Next.js apps your team won't dread
Structure, testing and the boring conventions that keep velocity high.